From 81f0cf1019cfe708943c80f42b3fc54e689b2273 Mon Sep 17 00:00:00 2001 From: Jannik Vogel Date: Sun, 20 Mar 2016 01:53:49 +0100 Subject: [PATCH] Rasterizer: Implement W-Buffer --- src/video_core/clipper.cpp | 4 +-- src/video_core/pica.h | 20 +++++++++++-- src/video_core/rasterizer.cpp | 28 +++++++++++++++++-- .../renderer_opengl/gl_rasterizer.cpp | 4 +-- 4 files changed, 45 insertions(+), 11 deletions(-) diff --git a/src/video_core/clipper.cpp b/src/video_core/clipper.cpp index 3d503486e..409b980ba 100644 --- a/src/video_core/clipper.cpp +++ b/src/video_core/clipper.cpp @@ -64,8 +64,6 @@ static void InitScreenCoordinates(OutputVertex& vtx) viewport.halfsize_y = float24::FromRaw(regs.viewport_size_y); viewport.offset_x = float24::FromFloat32(static_cast(regs.viewport_corner.x)); viewport.offset_y = float24::FromFloat32(static_cast(regs.viewport_corner.y)); - viewport.zscale = float24::FromRaw(regs.viewport_depth_range); - viewport.offset_z = float24::FromRaw(regs.viewport_depth_far_plane); float24 inv_w = float24::FromFloat32(1.f) / vtx.pos.w; vtx.color *= inv_w; @@ -78,7 +76,7 @@ static void InitScreenCoordinates(OutputVertex& vtx) vtx.screenpos[0] = (vtx.pos.x * inv_w + float24::FromFloat32(1.0)) * viewport.halfsize_x + viewport.offset_x; vtx.screenpos[1] = (vtx.pos.y * inv_w + float24::FromFloat32(1.0)) * viewport.halfsize_y + viewport.offset_y; - vtx.screenpos[2] = viewport.offset_z + vtx.pos.z * inv_w * viewport.zscale; + vtx.screenpos[2] = vtx.pos.z * inv_w; } void ProcessTriangle(const OutputVertex &v0, const OutputVertex &v1, const OutputVertex &v2) { diff --git a/src/video_core/pica.h b/src/video_core/pica.h index 4552ff81c..00c4fa79f 100644 --- a/src/video_core/pica.h +++ b/src/video_core/pica.h @@ -69,7 +69,7 @@ struct Regs { INSERT_PADDING_WORDS(0x9); BitField<0, 24, u32> viewport_depth_range; // float24 - BitField<0, 24, u32> viewport_depth_far_plane; // float24 + BitField<0, 24, u32> viewport_depth_near_plane; // float24 BitField<0, 3, u32> vs_output_total; @@ -121,7 +121,20 @@ struct Regs { BitField<16, 10, s32> y; } viewport_corner; - INSERT_PADDING_WORDS(0x17); + INSERT_PADDING_WORDS(0x1); + + //TODO: early depth + INSERT_PADDING_WORDS(0x1); + + INSERT_PADDING_WORDS(0x2); + + enum DepthBuffering : u32 { + WBuffering = 0, + ZBuffering = 1, + }; + BitField< 0, 1, DepthBuffering> depthmap_enable; + + INSERT_PADDING_WORDS(0x12); struct TextureConfig { enum WrapMode : u32 { @@ -1273,10 +1286,11 @@ ASSERT_REG_POSITION(cull_mode, 0x40); ASSERT_REG_POSITION(viewport_size_x, 0x41); ASSERT_REG_POSITION(viewport_size_y, 0x43); ASSERT_REG_POSITION(viewport_depth_range, 0x4d); -ASSERT_REG_POSITION(viewport_depth_far_plane, 0x4e); +ASSERT_REG_POSITION(viewport_depth_near_plane, 0x4e); ASSERT_REG_POSITION(vs_output_attributes[0], 0x50); ASSERT_REG_POSITION(vs_output_attributes[1], 0x51); ASSERT_REG_POSITION(viewport_corner, 0x68); +ASSERT_REG_POSITION(depthmap_enable, 0x6D); ASSERT_REG_POSITION(texture0_enable, 0x80); ASSERT_REG_POSITION(texture0, 0x81); ASSERT_REG_POSITION(texture0_format, 0x8e); diff --git a/src/video_core/rasterizer.cpp b/src/video_core/rasterizer.cpp index 5b9ed7c64..f5c6c9742 100644 --- a/src/video_core/rasterizer.cpp +++ b/src/video_core/rasterizer.cpp @@ -859,10 +859,31 @@ static void ProcessTriangleInternal(const Shader::OutputVertex& v0, } } + // interpolated_z = z / w + float interpolated_z_over_w = (v0.screenpos[2].ToFloat32() * w0 + + v1.screenpos[2].ToFloat32() * w1 + + v2.screenpos[2].ToFloat32() * w2) / wsum; + + // Z-Buffer (z / w * scale + offset) + float depth_scale = float24::FromRaw(regs.viewport_depth_range).ToFloat32(); + float depth_offset = float24::FromRaw(regs.viewport_depth_near_plane).ToFloat32(); + float depth = interpolated_z_over_w * depth_scale + depth_offset; + + // Potentially switch to W-Buffer + if (regs.depthmap_enable == Pica::Regs::DepthBuffering::WBuffering) { + float interpolated_w = interpolated_w_inverse.ToFloat32() * wsum; + + // W-Buffer (z * scale + w * offset = (z / w * scale + offset) * w) + depth *= interpolated_w; + + } + + // Clamp the result + depth = MathUtil::Clamp(depth, 0.0f, 1.0f); + + // Convert float to integer unsigned num_bits = Regs::DepthBitsPerPixel(regs.framebuffer.depth_format); - u32 z = (u32)((v0.screenpos[2].ToFloat32() * w0 + - v1.screenpos[2].ToFloat32() * w1 + - v2.screenpos[2].ToFloat32() * w2) * ((1 << num_bits) - 1) / wsum); + u32 z = (u32)(depth * ((1 << num_bits) - 1)); if (output_merger.depth_test_enable) { u32 ref_z = GetDepth(x >> 4, y >> 4); @@ -991,6 +1012,7 @@ static void ProcessTriangleInternal(const Shader::OutputVertex& v0, return 255 - combiner_output.a(); case Regs::BlendFactor::DestAlpha: + case Regs::BlendFactor::DestColor: return dest.a(); case Regs::BlendFactor::OneMinusDestAlpha: diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 6ca9f45e2..e6d5d3679 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -250,7 +250,7 @@ void RasterizerOpenGL::NotifyPicaRegisterChanged(u32 id) { // Depth modifiers case PICA_REG_INDEX(viewport_depth_range): - case PICA_REG_INDEX(viewport_depth_far_plane): + case PICA_REG_INDEX(viewport_depth_near_plane): SyncDepthModifiers(); break; @@ -865,7 +865,7 @@ void RasterizerOpenGL::SyncCullMode() { void RasterizerOpenGL::SyncDepthModifiers() { float depth_scale = -Pica::float24::FromRaw(Pica::g_state.regs.viewport_depth_range).ToFloat32(); - float depth_offset = Pica::float24::FromRaw(Pica::g_state.regs.viewport_depth_far_plane).ToFloat32() / 2.0f; + float depth_offset = Pica::float24::FromRaw(Pica::g_state.regs.viewport_depth_near_plane).ToFloat32() / 2.0f; // TODO: Implement scale modifier uniform_block_data.data.depth_offset = depth_offset;